Optimization of guide-vane-enhanced longitudinal vortex electrostatic precipitators based on artificial neural network and non-dominated sorting genetic algorithm for balancing particulate matter collection and energy consumption

Electrostatic precipitators (ESPs) face challenges in balancing fine particulate penetration and energy consumption, necessitating structural innovations for high-efficiency, low-resistance operation. This study proposes an optimized design using rectangular deflectors to actively induce longitudinal vortices within ESP channels, enhancing cross-flow particle migration-deposition while balancing collection efficiency (η) and pressure drop (ΔP). A validated multi-physics model coupling electric-flow-particle fields systematically analyzes geometric parameters (length “Lg,” height “Hg,” deflection angle “α,” and transverse plate spacing “Sg”). Leveraging Latin hypercube sampling (LHS) and an artificial neural network (ANN) surrogate model coupled with the NSGA-III algorithm, multi-objective optimization reveals a tradeoff: deflector-generated vortices increase mid-to-rear section efficiency by up to 18% versus baseline but nonlinearly amplify pressure drop. Since fan power consumption scales linearly with pressure drop for a fixed volumetric flow rate, additional energy consumption surges within high-efficiency operating zones. SHAP analysis identifies “Lg” as dominant for efficiency (52.5% contribution), while “Hg” and “Sg” primarily regulate pressure drop (48.6%/31.2%). The experimentally verified Pareto frontier offers engineering guidelines: prioritize “Lg” optimization with constrained “Hg” and “Sg” for efficiency focus, or strictly constrain “Hg” and “Sg” with coordinated tuning for pressure drop reduction. This work establishes a novel theoretical-practical framework for high-efficiency, low-resistance ESP design.